The proposed research program will serve complementary objectives in neuroscience and neural engineering. The neuroscience objective is to further knowledge of the role and nature of somatosensory feedback in multi-joint limb control. The engineering objective is to design a system for extracting limb-state information (e.g. limb position and velocity) from the firing rate modulations of primary afferent neurons. State feedback is required for closed-loop control of functional electrical stimulation systems, which are used to restore action to muscles paralyzed by spinal cord or other central nervous system injuries. To achieve these goals, we are using state-of-the-art technologies that enable large numbers of afferent neurons to be recorded simultaneously and chronically during natural motor behaviors such as standing, walking, and reaching. Multichannel recordings are essential for this work, because there are multiple degrees-of-freedom for joint movement and several different kinematic and kinetic state variables that are of interest for closed- loop control applications. These data also permit the direct examination of the role and nature of primary afferent neurons in the perception of body state known as proprioception which is formed through the integration of multiple primary afferent neuronal inputs. This proposal will achieve 4 Specific Aims, with each Aim being an incremental progression toward fulfilling the broader goals stated above. Specific Aim 1 is to quantify, using Information Theory, the limb-state information conveyed by single neurons under the following conditions: 1) intact spinal cord, 2) acutely transected spinal cord, and 3) chronically transected spinal cord. These results will be used to evaluate changes in the quality and reliability of information transmitted by primary afferent neurons before and after spinal cord injury, a condition known to cause plastic changes in spinal neuronal circuitry potentially altering the response properties of muscle spindles. Using a decerebate cat preparation, the following state variables for the hindlimb will be studied: position, velocity, acceleration, endpoint force (e.g. ground reaction force during stance), and stance phase joint-torque. Neural recordings will be made during limb movements imposed by a robot arm attached to the foot. The state variables will be measured in reference frames based in intrinsic, joint-space coordinates (i.e. intersegmental angles) or extrinsic, end point-space coordinates (i.e. Cartesian or polar coordinates for the toe relative to the hip) to compare the extent to which state-information carried by afferent neurons depends on the chosen frame of reference. Specific Aim 2 is to use the data collected in SA-1, to develop mathematical models to decode position, velocity, force, and torque information from the ensembles of simultaneously recorded afferent neurons. Specific Aim 3 is to decode in near real-time, limb-state from ensemble firing rates of afferent neurons recorded during hindlimb stepping movements evoked by passive movement and fixed-pattern electrical stimulation in intact spinal cord and chronically spinalized decerebrate cats. Specific Aim 4 is to implement online state-feedback decoding (SA-3) in a finite state control system for FES-evoked hindlimb stepping in intact spinal cord and chronically spinalized decerebrate cats.